v0.2.2

Tapir v0.2.2

Diff since v0.2.1

Merged pull requests:

• Fix small typo in README (#125) (@jacobusmmsmit)
• Update README.md (#126) (@willtebbutt)
• ADTypes Interop (#127) (@willtebbutt)

v0.2.1

Tapir v0.2.1

Diff since v0.2.0

Merged pull requests:

• LogDensityProblemsAD + fix CI bug (#123) (@willtebbutt)

Closed issues:

• Globals (#41)

v0.2.0

Tapir v0.2.0

Diff since v0.1.2

Merged pull requests:

• Update README.md (#119) (@willtebbutt)
• Wct/refactor api and ad implementation (#121) (@willtebbutt)

Closed issues:

• DifferentiationInterface supports Tapir! (#107)
• Tapir.value_and_pullback!! and mutation (#113)
• Split reverse mode for Tapir (#115)
• What happens to value_and_gradient!! if the output is not a Float64? (#116)
• is_primitive doesn't invalidate methods in TICache (#120)

v0.1.2

Tapir v0.1.2

Diff since v0.1.1

Merged pull requests:

• Refactor tangent testing utils (#109) (@willtebbutt)

v0.1.1

Tapir v0.1.1

Diff since v0.1.0

Merged pull requests:

• Remove redundant deps (#108) (@willtebbutt)
• Add pull request template (#110) (@willtebbutt)

v0.1.0

Tapir v0.1.0

Merged pull requests:

• CompatHelper: add new compat entry for ExprTools at version 0.1, (keep existing compat) (#1) (@github-actions[bot])
• Remove note on private umlaut fork (#4) (@yebai)
• Bug fixes and extra tests (#5) (@willtebbutt)
• Aliasing checks (#6) (@willtebbutt)
• Fix duplication problem (#9) (@willtebbutt)
• Fix setfield implementation in test utils (#10) (@willtebbutt)
• Incorporate DiffRules.jl in preparation for DiffTests.jl (#11) (@willtebbutt)
• WIP: optimisations (#12) (@willtebbutt)
• Improve tests + add new tests (#13) (@willtebbutt)
• Undefined array entries (#14) (@willtebbutt)
• Verify increment!! infers (#15) (@willtebbutt)
• jl_reshape_array and excluding memmove (#16) (@willtebbutt)
• unsafe_copyto! and improve test error messages (#17) (@willtebbutt)
• Improve testing (#18) (@willtebbutt)
• DiffTests integration (#19) (@willtebbutt)
• Small tweaks (#22) (@willtebbutt)
• Test all distributions (#23) (@willtebbutt)
• Support more distributions (#28) (@willtebbutt)
• Tackle array testing (#29) (@willtebbutt)
• Reorganise test structure for clarity (#34) (@willtebbutt)
• Update README.md (#36) (@willtebbutt)
• Make container-like testing the default (#37) (@willtebbutt)
• Extra setindex and getindex tests (#38) (@willtebbutt)
• Remove redundant rule (#39) (@willtebbutt)
• Basic tests on Turing.jl (#42) (@willtebbutt)
• Testing Improvements (#43) (@willtebbutt)
• Recursive Tape Structure (#44) (@willtebbutt)
• Towards Benchmarking: Attempt 2 (#47) (@willtebbutt)
• Benchmarking derived rules (#48) (@willtebbutt)
• Don't Unroll (#49) (@willtebbutt)
• Simplify Reverse Pass (#50) (@willtebbutt)
• Turing + Performance Fixes (#51) (@willtebbutt)
• General Tidy Up (#52) (@willtebbutt)
• Begin improving allocation-free testing (#53) (@willtebbutt)
• Simple Interface (#54) (@willtebbutt)
• Improve setfield performance (#55) (@willtebbutt)
• Tidy up and improve tangent_type performance (#56) (@willtebbutt)
• Update Turing tests and fix intermittent benchmarking failures (#57) (@willtebbutt)
• Produce performance histogram (#59) (@willtebbutt)
• Systematic composite type generation and testing (#60) (@willtebbutt)
• Enable a range of additional tests (#61) (@willtebbutt)
• Include KernelFunctions and AbstractGPs in the integration tests (#62) (@willtebbutt)
• Uncomment tests (#63) (@willtebbutt)
• Update README.md (#64) (@willtebbutt)
• Do static dispatch properly (#65) (@willtebbutt)
• More Turing.jl integration tests (#66) (@willtebbutt)
• Wct/minor tweaks (#67) (@willtebbutt)
• Fix up benchmarking (#70) (@willtebbutt)
• More formatting improvements to benchmarking (#71) (@willtebbutt)
• Static is_primitive (#72) (@willtebbutt)
• More cosmetic stuff (#73) (@willtebbutt)
• Replace Vector stack with Stack stack (#74) (@willtebbutt)
• Tidy up typeof and friends (#75) (@willtebbutt)
• Stacks Not Slots (#76) (@willtebbutt)
• Only use stacks for tangents (#77) (@willtebbutt)
• Simplify PhiNodes (#78) (@willtebbutt)
• Optimise for NoTangent (#79) (@willtebbutt)
• Cache Rules in AbstractInterpreter (#80) (@willtebbutt)
• Dynamic Info Bug (#81) (@willtebbutt)
• Routinely comparing performance between frameworks (#83) (@willtebbutt)
• Fix Recursion and add test (#84) (@willtebbutt)
• Improve Small-Union Performance (#85) (@willtebbutt)
• Add dependabot (#86) (@willtebbutt)
• Add comment on test (#87) (@willtebbutt)
• ChainRules Macro (#88) (@willtebbutt)
• Add the Turing broadcasting example to the inter-framework benchmarks (#89) (@willtebbutt)
• Source-to-source (#90) (@willtebbutt)
• Use 1.10 in CI (#91) (@willtebbutt)
• CompatHelper: add new compat entry for Graphs at version 1, (keep existing compat) (#96) (@github-actions[bot])
• Fix interface (#98) (@willtebbutt)
• Set version (#99) (@willtebbutt)
• Add MIT license (#100) (@willtebbutt)
• Update README.md (#101) (@willtebbutt)
• Complete Phi rename (#102) (@willtebbutt)
• Remove out of date comment in docstring (#103) (@willtebbutt)
• Rename Phi to Tapir throughout package (#104) (@willtebbutt)

Closed issues:

• CI tests are broken (#2)
• Understanding performance (#3)
• Integration tests against DiffTests (#7)
• Distributions.jl tests (#20)
• Write input-validation tests (#21)
• Practicalities of Incorporating Existing ChainRules (#24)
• ReverseDiff limitations and issues (#25)
• More restrictive use of the rrule name (#26)
• Tests against all Array/AbstractArray APIs (#27)
• Improve missing foreigncall error message (#30)
• Add tests for KernelFunctions and AbstractGPs. (#45)
• Update README to include source-2-source and interpreter modes for Taped (#92)
• Array output seems to be initialized but not filled (#97)